The purpose of this study is to assess predictive value of the positron emission tomography (PET) with 18F-fluoro-deoxyglucose (FDG) for recurrence and survival after radiotherapy (RT) for non-small cell lung cancer (NSCLC). One hundred forty-nine patients underwent pretreatment PET (n = 67) or PET/computed tomography (CT) (n = 82) and definitive RT for NSCLC. We evaluated the relationship between the maximum-pixel standardized uptake value (SUV(max)) and clinical tumor features. Univariate Cox proportional hazard analysis (UVA) was used to quantify the risk for local-regional recurrence, distant metastases, and death. Multivariate Cox proportional hazard analysis (MVA) was used to assess the potential independent effect of SUV(max). In the PET group, T1 tumors showed significantly lower SUV(max) than T2, T3, and T4 tumors; in the PET/CT group, T1 tumors showed significantly lower SUV(max) than T3 and T4 tumors. A high SUV(max )was a negative factor for local-regional control (LRC) (p < 0.001), distant metastasis-free survival (DMFS) (p = 0.02), and overall survival (OS) (p = 0.001) on UVA in the PET group. However, the significance decreased to 0.05 for LRC, 0.04 for DMFS, and 0.04 for OS by MVA when tumor size was included in the analysis. A high SUV(max) was not a negative factor for LRC, DMFS, or OS on UVA and MVA in the PET/CT group. In conclusion, assessment of predictive value of SUV(max) for NSCLC requires consideration of primary tumor size, and the evidence is not sufficient to suggest that FDG uptake in a primary NSCLC provides prognostic information.